Data Center Availability: Why Redundancy is Not the Same as Continuity
Updated: Nov 5

One failed host can affect several applications when multiple virtual machines share the same physical server.
Data center availability is usually discussed in terms of spare capacity: a second host, a second power feed, a second storage path. Those matter, but they leave out what the applications on the failed host experience.
Redundancy answers one question, continuity another
Redundancy asks whether another resource is available if the first one fails. Continuity asks what happens to the application and its users while the workload moves to it.
Take a host carrying an application server, its database and a monitoring tool. In a recovery-based High Availability setup, all three may need to restart elsewhere. The application may come up before its database is ready, users may need to log in again, and queued requests may need to be processed or checked before anyone trusts the result. None of this means data is corrupted or transactions are lost; that depends on the application, the storage design and what was in progress at the moment of failure.
For many workloads, that recovery is perfectly acceptable, and High Availability is the right design.
When the restart itself is the problem
Some workloads cannot comfortably absorb a restart: a time-sensitive process is interrupted, an operational function disappears, or the checking afterwards is substantial. For those, Fault Tolerance is worth evaluating.
vServerFT synchronises CPU execution, memory and disk state in real time across two physical nodes, and is designed to keep the protected workload running through a covered hardware failure without requiring the application to restart.
The protection applies to the workloads placed in the FT environment, not to every service in the data center. Storage, network paths, external services and application-level faults still need their own controls. For a side-by-side view of the two approaches, see High Availability vs Fault Tolerance, and for the underlying argument, Why Fault Tolerance Matters: Real Benefits Beyond Failover.
In practice, the useful exercise is to list the virtual machines on a host and ask what a restart would cost each one. Confirm that the application vendor supports the proposed virtual environment, then test a host failure and watch what users and dependent systems actually experience, including failures outside the FT platform.




